基于Matlab的移动物体检测与识别

发布时间：2018-05-21 13:46

本文选题：移动物体检测 + 无人机　；参考：《西安石油大学》2017年硕士论文

【摘要】：随着国内外计算机技术的快速发展,移动物体检测技术在计算机视觉和人工智能上占有非常重要的地位,各种移动物体检测算法适用于不同简单或复杂环境,特别是在最近几年无人驾驶与人机交互上的应用。本课题的应用背景基于无人机自动返航过程中受GPS给定精度限制,导致无人机降落误差半径3米,提出使用移动物体检测技术解决上述限制,并仿真验证了利用移动物体检测技术的可行性。本文首先以追踪和确定无人机在视频中的位置信息为背景,介绍和比较了三种常用的利用帧间关系移动检测算法的原理和综合性能,并对效果最好的背景减除算法的参数选取给出了明确的方法,然后对无人机检测的干扰进行滤除,在Matlab软件上进行移动检测仿真,实验结果表明在三种方法的计算时间和效果上,背景减除算法性能最好,且它的前景模板基本表示了真实移动目标。其次本文在背景减除算法的基础上,对检测出来的前景模板进行矩形框标记,并计算出了图像中无人机中心与图像中心的像素距离和方位角,利用矩形框的面积与无人机的实际高度进行了简单标定。调试结果表明,高度越低时,无人机的中心与图像中心的距离和矩形框的面积越准确。
[Abstract]:With the rapid development of computer technology at home and abroad, mobile object detection technology plays a very important role in computer vision and artificial intelligence. Especially in recent years in the application of unmanned and human-computer interaction. The application background of this subject is based on the limitation of GPS precision in the automatic return of UAV, which leads to the error radius of UAV landing 3 meters. This paper proposes to use mobile object detection technology to solve the above limitation. The feasibility of using moving object detection technology is verified by simulation. Firstly, based on tracking and determining the position information of UAV in video, this paper introduces and compares the principle and comprehensive performance of three commonly used motion detection algorithms using inter-frame relationship. And the best background subtraction algorithm parameters selection method is given, then the UAV detection interference is filtered, and the simulation of mobile detection is carried out on the Matlab software. The experimental results show that the performance of the background subtraction algorithm is the best in terms of the computation time and effect of the three methods, and its foreground template basically represents the real moving target. Secondly, on the basis of background subtraction algorithm, the detected foreground template is marked by rectangular frame, and the pixel distance and azimuth between UAV center and image center are calculated. The area of rectangular frame and the actual height of UAV are simply calibrated. The test results show that the distance between the center of UAV and the image center and the area of rectangular frame are more accurate when the height is lower.
【学位授予单位】：西安石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】